Conversion provides several advantages for gene mapping projects of complex diseases such as cancer. The approach takes advantage of selective retention of a subset of human chromosomes within somatic cell hybrids, isolating single copies of all desired human chromosome pairs. This strategy functions both to simplify mutation detection and to allow unambiguous phase information to be determined. Constructing haplotypes on the basis of conventional genotype and pedigree data is challenging, particularly for diseases with a late age of onset, such as cancer.

The homologue retained in a particular hybrid is identified by conventional genotyping of a few markers per chromosome. These “haploid” hybrids can then be used to increase the sensitivity of traditional mutation analysis, because the disease-causing chromosome will not be accompanied by the normal wild-type allele. Haplotypes may also be determined simply by genotyping each haploid hybrid. We have typed DNA from 100 hybrids to examine chromosomal retention patterns, test the feasibility of conversion for analysis of whole genomes, and evaluate assumptions regarding appropriate experimental design.

We have investigated the theoretical efficiency of using haplotypes compared with conventional genotypes in linkage and linkage disequilibrium studies. In the linkage disequilibrium setting, we determined the Fisher information (with respect to haplotype frequency) provided by unrelated individuals. For haplotypes with two single-nucleotide polymorphisms, conversion provides 5–45% more information per subject than standard genotyping, depending on true haplotype frequencies; for haplotypes with five single-nucleotide polymorphisms, improvement ranges from 20% to 92%. Since Fisher information is inversely related to sample size, conversion requires up to 12.5 times fewer subjects than standard genotyping to obtain the same information. We are currently completing similar calculations in the linkage analysis setting.

The extra cost associated with conversion includes hybrid construction and characterization, and duplicate genotyping. However, the increased information per subject reduces overall recruitment and phenotyping costs, which tend to exceed those for genotyping. The efficiency of using conversion to construct haplotypes will probably increase even further as automated methods of genotyping continue to improve.